triangulate
Skill1 starsChanged 2 months agoMITView on GitHub
Triangulate plans and code from three expert viewpoints — functionality, security, and testing — across three phases (plan, implementation, review). Each iteration sharpens the position of remaining issues until all are resolved. Always use this skill when: asked to review plans, code, or branches; asked to evaluate from functionality/security/testing perspectives; asked for PR or pre-implementation review; asked to implement or develop from a plan.
3 other projects use a version of this.
Adopted by
Most recently taken up by
- sophonfinance-wq/finance-automation-portfoliotheir own version, first seen Nov 1, 2026
- Adham-Aly/triangulatetheir own version, first seen Sep 27, 2026
- vasilievyakov/triangulate-skilltheir own version, first seen Jan 13, 2026
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- sophonfinance-wq/finance-automation-portfolio★ 16same name
- vasilievyakov/triangulate-skill★ 9same name
- Adham-Aly/triangulate★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository10 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- global/CLAUDE.mdCLAUDE.md
- agent-reviewSkill
- context-budgetSkill
- exploreSkill
- pr-createSkill
- retrospectSkill
- security-scanSkill
- simplifySkill
- test-genSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 0 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 29
Read phase files and common-rules.digest.md with the Read tool, whole-file.
- Very specific: line 30
Every phase file's last line is ## END-OF-PHASE-<N>, and common-rules.digest.md's last line is ## END-OF-DIGEST.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Triangulate Skill
A skill that covers the entire development workflow from plan creation to coding to code review. Three expert agents (functionality, security, testing) triangulate issues at each phase, repeating review and fix cycles until all findings are resolved.
The skill is split across several files for context efficiency. Load only the files required for the current phase.
Supplemental Files
| File | Load when |
phases/phase-1-plan.md | Plan creation / plan review is the active phase |
phases/phase-2-coding.md | Implementation is the active phase |
phases/phase-3-review.md | Code review is the active phase |
common-rules.digest.md | Any phase — compact routing index for recurring rules (R1-R57, RS*, RT*); read before selecting rules |
common-rules.md | Targeted lookup only — full rule rows, extended obligations, severity tables, and shared orchestration obligations |
Loading protocol — Read common-rules.digest.md first. Match its pattern names against the diff and task, then use anchored rg queries to extract only the triggered rows from common-rules.md; also extract a selected rule's Extended obligations section when its row points there. For a named non-recurring Common Rules section, extract that heading and its bounded section. Read the full common-rules.md only when targeted extraction is inconclusive, and record the reason. Do not paraphrase rule details from memory.
Truncation protocol — a partial read of a phase file is silent unless you check for it:
- Read phase files and
common-rules.digest.mdwith theReadtool, whole-file. Do notcat/headthem through Bash — Bash output passes through an output-compressing proxy that can render a 500-line file as its first line plus a[N more lines]marker. - Every phase file's last line is
## END-OF-PHASE-<N>, andcommon-rules.digest.md's last line is## END-OF-DIGEST. If the last line you received is not that terminator, the read was partial — re-read before acting on it. - Before reporting a phase complete, reconcile the steps you executed against the
step_ids:declared in that phase's front matter. An unexecuted ID means the phase is not complete. The step named incore:has no substitute — inline work by the orchestrator does not discharge it.
<!-- Machine-parsed by hooks/check-rule-sync.sh (check 8) BY SHAPE: backticked
key tokens, and a backticked stem followed by "(phase files)" / "(the
digest)". Reword the sentence and the stems stop extracting. -->Manifest keys: step_ids:, core:. Terminator stems: END-OF-PHASE (phase files), END-OF-DIGEST (the digest).
Entry Point Decision
Determine the starting phase from the user's instructions:
| User instruction | Starting phase | First action |
| "Implement", "Develop", etc. — starting from scratch | Phase 1 (Plan creation) | Read phases/phase-1-plan.md |
| An existing plan file path is specified | Phase 1 (From review) | Read phases/phase-1-plan.md |
| "Review the code", "Review the branch" | Phase 3 (Code review) | Read phases/phase-3-review.md |
Phase Transitions
- Phase 1 → Phase 2: after plan review completes, Read
phases/phase-2-coding.md - Phase 2 → Phase 3: after coding completes, Read
phases/phase-3-review.md - Phase 3 standalone: when invoked for branch review only, skip Phase 1 and 2
Each phase file ends with a summary, a pointer to the next phase, and its ## END-OF-PHASE-N terminator as the final line. Follow the pointer when the current phase reports complete.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.